A First-Order Algorithmic Framework for Wasserstein Distributionally Robust Logistic Regression

被引:0
作者
Li, Jiajin [1 ]
Huang, Sen [1 ]
So, Anthony Man-Cho [1 ]
机构
[1] Chinese Univ Hong Kong, Dept Syst Engn & Engn Management, Shatin, Hong Kong, Peoples R China
来源
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 32 (NIPS 2019) | 2019年 / 32卷
关键词
CONVEX-OPTIMIZATION;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Wasserstein distance-based distributionally robust optimization (DRO) has received much attention lately due to its ability to provide a robustness interpretation of various learning models. Moreover, many of the DRO problems that arise in the learning context admits exact convex reformulations and hence can be tackled by off-the-shelf solvers. Nevertheless, the use of such solvers severely limits the applicability of DRO in large-scale learning problems, as they often rely on general purpose interior-point algorithms. On the other hand, there are very few works that attempt to develop fast iterative methods to solve these DRO problems, which typically possess complicated structures. In this paper, we take a first step towards resolving the above difficulty by developing a first-order algorithmic framework for tackling a class of Wasserstein distance-based distributionally robust logistic regression (DRLR) problem. Specifically, we propose a novel linearized proximal ADMM to solve the DRLR problem, whose objective is convex but consists of a smooth term plus two non-separable non-smooth terms. We prove that our method enjoys a sublinear convergence rate. Furthermore, we conduct three different experiments to show its superb performance on both synthetic and real-world datasets. In particular, our method can achieve the same accuracy up to 800+ times faster than the standard off-the-shelf solver.
引用
收藏
页数:11
相关论文
共 28 条
  • [11] Gao R, 2017, PREPRINT
  • [12] First-Order Algorithms for Convex Optimization with Nonseparable Objective and Coupled Constraints
    Gao X.
    Zhang S.-Z.
    [J]. Zhang, Shu-Zhong (zhangs@umn.edu), 1600, Springer Science and Business Media Deutschland GmbH (05): : 131 - 159
  • [13] On the linear convergence of the alternating direction method of multipliers
    Hong, Mingyi
    Luo, Zhi-Quan
    [J]. MATHEMATICAL PROGRAMMING, 2017, 162 (1-2) : 165 - 199
  • [14] Lee C., 2015, DISTRIBUTIONAL UNPUB
  • [15] A MAJORIZED ADMM WITH INDEFINITE PROXIMAL TERMS FOR LINEARLY CONSTRAINED CONVEX COMPOSITE OPTIMIZATION
    Li, Min
    Sun, Defeng
    Toh, Kim-Chuan
    [J]. SIAM JOURNAL ON OPTIMIZATION, 2016, 26 (02) : 922 - 950
  • [16] A HIGHLY EFFICIENT SEMISMOOTH NEWTON AUGMENTED LAGRANGIAN METHOD FOR SOLVING LASSO PROBLEMS
    Li, Xudong
    Sun, Defeng
    Toh, Kim-Chuan
    [J]. SIAM JOURNAL ON OPTIMIZATION, 2018, 28 (01) : 433 - 458
  • [17] Lin Z. C., 2011, P 24 INT C NEUR INF, P612
  • [18] Decomposition algorithm for distributionally robust optimization using Wasserstein metric with an application to a class of regression models
    Luo, Fengqiao
    Mehrotra, Sanjay
    [J]. EUROPEAN JOURNAL OF OPERATIONAL RESEARCH, 2019, 278 (01) : 20 - 35
  • [19] Namkoong H, 2016, ADV NEUR IN, V29
  • [20] Shafieezadeh-Abadeh S, 2015, ADV NEUR IN, V28